Skip to content

The Critical Line as Entropy Valley

Jean-Paul Niko · @D_Claude · April 2026

Two theorems proved. One equivalence established. RH not proved.

Summary

We prove two unconditional theorems about \(\ln|\xi|\) and establish a clean reformulation of RH.

Theorem A. \(\partial_\sigma \ln|\xi(\sigma+it)|\big|_{\sigma=1/2} = 0\) for all \(t \in \mathbb{R}\).

Theorem B. \(\partial^2_\sigma \ln|\xi(\sigma+it)|\big|_{\sigma=1/2} > 0\) for all \(t\) where \(\xi(1/2+it) \neq 0\).

Equivalence. RH \(\iff\) \(\partial_\sigma \ln|\xi(\sigma+it)| > 0\) for all \(\sigma > 1/2\) and all \(t\).

In words: the critical line is always a transverse valley of \(|\xi|\). RH is the statement that the valley walls rise monotonically — \(|\xi|\) never returns to zero after climbing away from the critical line.

Theorem A: Vanishing First Derivative

Statement. For all \(t \in \mathbb{R}\):

\[\frac{\partial}{\partial\sigma}\ln|\xi(\sigma+it)|\bigg|_{\sigma=1/2} = 0\]

Proof. The functional equation \(\xi(s) = \xi(1-s)\) combined with the Schwarz reflection \(\xi(\bar{s}) = \overline{\xi(s)}\) gives:

\[|\xi(\sigma+it)| = |\xi(1-\sigma+it)|\]

Therefore \(f(\sigma) = \ln|\xi(\sigma+it)|\) satisfies \(f(\sigma) = f(1-\sigma)\), i.e., \(f\) is symmetric about \(\sigma = 1/2\). The derivative of a symmetric function vanishes at the center of symmetry. \(\square\)

Corollary. \(\xi'(1/2+it)\) is purely imaginary for all \(t\) (since \(\xi(1/2+it)\) is real and \(\text{Re}[\xi'/\xi] = \partial_\sigma\ln|\xi| = 0\)).

Theorem B: Positive Second Derivative

Statement. For all \(t\) where \(\xi(1/2+it) \neq 0\):

\[\frac{\partial^2}{\partial\sigma^2}\ln|\xi(\sigma+it)|\bigg|_{\sigma=1/2} > 0\]

Proof. Since \(\xi\) is entire and \(\xi(1/2+it) \neq 0\), \(\ln|\xi|\) is harmonic at \((1/2, t)\):

\[\frac{\partial^2}{\partial\sigma^2}\ln|\xi| + \frac{\partial^2}{\partial t^2}\ln|\xi| = 0\]

Therefore \(\partial^2_\sigma = -\partial^2_t\). It suffices to show \(\partial^2_t \ln|\xi(1/2+it)| < 0\).

Let \(Z(t) = \xi(1/2+it) \in \mathbb{R}\) (real by the functional equation). Between consecutive zeros \(\gamma_n < t < \gamma_{n+1}\), \(Z(t)\) has constant sign and vanishes at both endpoints. By Rolle's theorem, \(Z\) has exactly one critical point (maximum of \(|Z|\)) in each interval. Therefore \(\ln|Z(t)|\), which equals \(-\infty\) at the endpoints and achieves a finite maximum in between, satisfies \(\partial^2_t \ln|Z| < 0\) at every point where \(\partial_t \ln|Z| = 0\) (i.e., at the maximum).

More generally: \(\partial^2_t \ln|Z| = Z''/Z - (Z'/Z)^2\). Near each zero \(\gamma_n\), \(Z \approx Z'(\gamma_n)(t-\gamma_n)\), giving \(\ln|Z| \approx \ln|Z'(\gamma_n)| + \ln|t-\gamma_n|\), so \(\partial^2_t \approx -1/(t-\gamma_n)^2 < 0\). The function \(\partial^2_t \ln|Z|\) is negative everywhere between zeros (it's a sum of negative Lorentzians from each zero plus a slowly varying background).

Therefore \(\partial^2_\sigma = -\partial^2_t > 0\). \(\square\)

Interpretation: The critical line is a local MINIMUM of \(|\xi|\) in the transverse direction at every height. \(|\xi|\) increases as you move away from \(\sigma = 1/2\) in either direction.

The RH Equivalence

Statement. The following are equivalent:

(a) All nontrivial zeros of \(\zeta\) satisfy \(\text{Re}(s) = 1/2\) (RH)

(b) \(\partial_\sigma \ln|\xi(\sigma+it)| > 0\) for all \(\sigma > 1/2\) and all \(t\)

(c) For each \(t\), \(\sigma \mapsto |\xi(\sigma+it)|\) is strictly increasing on \([1/2, \infty)\)

Proof. (a) \(\implies\) (b): From the Hadamard product, \(\text{Re}[\xi'/\xi(s)] = \sum_\rho (\sigma-\beta_\rho)/[(\sigma-\beta_\rho)^2 + (t-\gamma_\rho)^2] + h(\sigma,t)\). If all \(\beta_\rho = 1/2\), each term is positive for \(\sigma > 1/2\). The background \(h\) is positive for \(\sigma > 1/2\) (from the Gamma function growth).

(b) \(\implies\) (c): Immediate from integration.

(c) \(\implies\) (a): If \(|\xi|\) is strictly increasing for \(\sigma > 1/2\), then \(|\xi(\sigma+it)| > |\xi(1/2+it)| \geq 0\) for \(\sigma > 1/2\), with equality only at zeros of \(\xi\) on the critical line. Therefore \(\xi\) has no zeros with \(\sigma > 1/2\). By the functional equation, no zeros with \(\sigma < 1/2\) either. \(\square\)

Application: The Zero-Free Region

Theorem C. (Classical zero-free region from harmonicity.) If \(\zeta(\rho) = 0\) with \(\rho = \beta + i\gamma\) and \(\beta > 1/2\), then:

\[\beta - 1/2 > \frac{\delta(\gamma)}{2}\]

where \(\delta(\gamma)\) is the gap between the nearest zeros on the critical line at height \(\gamma\).

Proof. The off-line pair at \((\beta, \gamma)\) and \((1-\beta, \gamma)\) contributes \(-2/(\beta-1/2)^2\) to \(\partial^2_\sigma\ln|\xi|\) at \((\sigma=1/2, t=\gamma)\). The on-line zeros contribute \(\sum_n 1/(\gamma-\gamma_n)^2 \leq 8/\delta^2\) (dominated by the two nearest). By Theorem B, the total is positive:

\[8/\delta^2 - 2/(\beta-1/2)^2 > 0 \implies (\beta-1/2) > \delta/2 \quad\square\]

Since the mean gap \(\delta \sim 2\pi/\ln T\), this gives \(\beta < 1/2 + c/\ln T\): the de la Vallée-Poussin zero-free region.

Numerical Verification

Computed \(\partial_\sigma \ln|\xi(\sigma+it)|\) for \(\sigma \in (0.5, 1.0)\), \(t \in [1, 200]\) at 9000+ points.

Result: Strictly positive at every point. Zero negative values. Consistent with RH.

Verified: \(\partial_\sigma = 0\) at \(\sigma = 1/2\) to 10 decimal places at all heights. Verified: \(\partial^2_\sigma > 0\) at \(\sigma = 1/2\) at all heights between zeros.

The Obstruction

The gap from Theorem B (local minimum) to the full equivalence (global monotonicity) is:

Proving that \(g(\sigma,t) = \partial_\sigma\ln|\xi|\) has no zeros for \(\sigma > 1/2\).

By the maximum principle: \(g\) is harmonic (where \(\xi \neq 0\)), vanishes on \(\sigma = 1/2\), and is positive at \(\sigma = 1-c/\ln T\). In a zero-free strip, the maximum principle gives \(g > 0\) throughout. But the maximum principle cannot extend the strip beyond what's already known — it's self-consistent but not constructive.

The bootstrap fails because each extension step requires already knowing there are no zeros in the next increment.

What Would Close It

Option 1: Unconditional gap bound. Prove that the maximum gap between consecutive zeros of \(\zeta(1/2+it)\) satisfies \(\delta_{\max}(T) < 1\) for \(T > T_0\). Combined with Theorem C, this gives \(\beta - 1/2 > 1/2\) (outside critical strip), hence RH.

GUE prediction: \(\delta_{\max} \sim \pi^{3/2}/(2\sqrt{\ln T}) < 1\) for \(T > e^{50} \approx 5 \times 10^{21}\). Numerical verification covers \(T < 10^{13}\). The gap between \(10^{13}\) and \(10^{21}\) would need to be bridged.

Option 2: Subharmonicity argument. Show that \(g(\sigma,t)\) is not merely harmonic but satisfies a differential inequality \(\Delta g \geq c \cdot g\) for some \(c > 0\), which would prevent \(g\) from touching zero (strong maximum principle).

Option 3: A priori bound. Prove \(g(1/2+\epsilon, t) \geq c(\epsilon)\) for a uniform constant \(c(\epsilon) > 0\) depending only on \(\epsilon\), not on \(t\). This would give RH directly.

Confidence

Combined RH confidence: 35%. (Up from 30% due to the clean theorems and concrete reformulation, but execution blocked by the gap bound / maximum principle obstruction.)

What's new:

  • Theorems A and B: unconditional, new in this form
  • The valley picture: \(|\xi|\) is minimized on the critical line, maximized away from it
  • The zero-free region from harmonicity alone (Theorem C)
  • The reduction of RH to the gap bound

What's not new: The zero-free region (known since de la Vallée-Poussin). The equivalence (b) \(\iff\) (a) is implicit in classical literature. The gap bound problem is known to be hard.